driesknapen . net

Update: Much to my pleasure, it seems that all of these issues have been addressed in iWork '09! Thanks, Apple :)

When I was watching Tuesday's Apple Special Event I was pleased to see that the rumours about a spreadsheet application making its way into iWork were true. I'm a heavy Keynote user - I completely abandoned Powerpoint since I switched to the Mac about two years ago - and I really like its ease of use and the way it's integrated with other applications.

While I was eagerly waiting for the 470MB 30 day trial version of iWork to download onto my Desktop, I read some early reviews on Numbers (which is the name of Apple's new spreadsheet application). They all talk about "the Apple way of doing spreadsheets", and focus on Numbers' great graphical and layout capabilities, easy-to-use interface and Excel import function.

However, I quickly began to fear that the few spreadsheet features that I rely on every day as a scientist were not going to be there. Sure, the software looks and feels great, and I appreciate that Numbers hasn't been designed with a scientist's requirements in mind. I don't use Excel for anything else than data storage, preliminary exploration and data preparation before using dedicated statistical packages to perform the actual analyses. So with just a few extra features Numbers would have managed to entirely replace Excel on my computer...

A few examples should make this clear (screenshots included):

• regression lines
  I can be brief about this one: Numbers doesn't support regression analyses, including regression line fitting (called trendlines in Excel) and R² calculations. For me, this is a deal breaker as I use this feature every day. I imported a complex and large Excel file containing dozens of worksheets and graphs, and although all worksheets and graphs were properly recognized, the regression lines and regression data from Excel were removed completely without any mention of it in the long import warnings list.

• error bars
  Similarly, Numbers doesn't support error bars in its graphs (although it contains formulae to calculate standard deviations and variances). A graph with error bars imported from Excel looses them, but this is mentioned in the import warning list. Funny though that during this import, the warning also contains something about trendlines, whereas in the example where I actually imported a trendline, it didn't...

• reuse of data within the same graph
  In a scatter plot (I haven't tried other graph types), you can't use the same data twice in the same graph. Suppose you have one column with the date, a second with income on that date, and a third one with expenses. Obviously you want to use the first column data twice as the x-axis, and use the two other columns to define two different data series. With Numbers, you have to copy the first column data over to a new column, because as soon as that data is part of a graph, it can no longer be used within the same graph.

• support for logarithmic axes
  This option is available only for 2D bar/column charts that aren’t stacked, and for line charts. So no logarithmic axes for scatter plots, where I tend to use them most. As a test, I created an Excel file with a simple scatter plot having one logarithmic axis and imported it into Numbers, and the logarithmic axis had been reset to a linear one. The import warnings mentioned that the minimum and maximum axis values had been recalculated and that my graph may look different. Sure enough, the graph showed a simple linear plot.

There are some other small annoyances that I encountered, such as the fact that the graph legend is a separate object and doesn't move along with the graph, but those are details. The lack of proper data exploration tools however made me decide to stick with Excel for now, unfortunately.